> For non-quantum objects, this isn’t a problem, as different methods of measuring an object all give you the same answer. Whether you use a measuring stick (like a ruler), high-definition imaging, or a physics-reliant technique like Brownian motion or gravitational settling, you’ll arrive at identical solutions.
Right off the top, it's made clear this author doesn't know what he or she is talking about. From at least the time of de Moivre, around 1740, the problem of estimating an accurate true value from multiple differing measurements was recognized in its own right as a problem. By 1810, Gauss and Laplace had discovered the basics of statistical estimation, the form of the normal distribution, results including the central limit theorem, and regression techniques such as non-linear weighted least squares. Now, almost 200 years later, the problem still exists, and is being worked on in various forms and circumstances.
To give a simple example, take 1000 frames of well-exposed video of a still scene, with constant illumination, and constant exposure settings on the camera. Now try to estimate how many photons, within a constant multiplicative factor, are coming from each point in the scene. Most of the measurements will show a normal distribution, but at the extremes of exposure, something more complex is going on (perhaps to do with sensor non-linearity, censored sampling, or something else). This is one of the basic theoretical problems in HDR imaging, and is an active area research (i.e. unsolved, as of today).
It's very hard to get past that paragraph -- which is the first one in answer to the question "why?".